Source-free Domain Adaptation Requires Penalized Diversity

TL;DR

This paper introduces a novel source-free domain adaptation method called Penalized Diversity (PD), which promotes representational diversity using separate feature extractors and mitigates weak hypotheses with regularization, improving adaptation across various domains.

Contribution

The paper proposes a new unsupervised SFDA algorithm that combines distinct backbone architectures and a regularizer to enhance diversity and robustness in domain adaptation.

Findings

PD outperforms existing methods on natural, synthetic, and medical datasets.

The combination of DBA and WHP improves adaptation under covariate and label shifts.

Empirical results show significant gains in predictive performance across diverse domain shifts.

Abstract

While neural networks are capable of achieving human-like performance in many tasks such as image classification, the impressive performance of each model is limited to its own dataset. Source-free domain adaptation (SFDA) was introduced to address knowledge transfer between different domains in the absence of source data, thus, increasing data privacy. Diversity in representation space can be vital to a model`s adaptability in varied and difficult domains. In unsupervised SFDA, the diversity is limited to learning a single hypothesis on the source or learning multiple hypotheses with a shared feature extractor. Motivated by the improved predictive performance of ensembles, we propose a novel unsupervised SFDA algorithm that promotes representational diversity through the use of separate feature extractors with Distinct Backbone Architectures (DBA). Although diversity in feature space is increased, the unconstrained mutual information (MI) maximization may potentially introduce amplification of weak hypotheses. Thus we introduce the Weak Hypothesis Penalization (WHP) regularizer as a mitigation strategy. Our work proposes Penalized Diversity (PD) where the synergy of DBA and WHP is applied to unsupervised source-free domain adaptation for covariate shift. In addition, PD is augmented with a weighted MI maximization objective for label distribution shift. Empirical results on natural, synthetic, and medical domains demonstrate the effectiveness of PD under different distributional shifts.